#
#asahi-gpu · 2023-08-11 13:21
linaand --loop=300 to see the last frame more
#asahi-gpu · 2023-08-11 13:21
linaUmm try --sb
#asahi-gpu · 2023-08-11 13:21
alyssai'm not seeing anything
#asahi-gpu · 2023-08-11 13:21
alyssahow do I use this apitrace?
#asahi-gpu · 2023-08-11 13:20
linaThe FS side manifests in do_clip() and happens even if all paths there return good near-1.0 alpha values (but not if all return 1.0 because that causes everything to be optimized out)
#asahi-gpu · 2023-08-11 13:19
linaFrom which I deduce that cmpsel stuff propagates NaNs? If not, something else is even more wrong...
#asahi-gpu · 2023-08-11 13:19
linaThere whole thing seems to spill into the FS, my guess is NaN propagation? That's the only thing that makes sense since the FS breaks even when all code paths must return good values.
#asahi-gpu · 2023-08-11 13:18
linaIn the VS, messing with fetch_clip_area causes it to work or not, seemingly depending on the order/polarity of the condition there
#asahi-gpu · 2023-08-11 13:18
alyssastill waking up
#asahi-gpu · 2023-08-11 13:18
alyssatbd if I get to it today
#asahi-gpu · 2023-08-11 13:18
alyssaspicy
#asahi-gpu · 2023-08-11 13:18
linaThe bad draw is 6918
#asahi-gpu · 2023-08-11 13:17
linaalyssa: https://cloud.asahilina.net/s/omCgkcc9LAHKjte
#asahi-gpu · 2023-08-11 13:11
alyssacould be an RA bug or something
#asahi-gpu · 2023-08-11 13:11
alyssayeah, send the apitrace I guess
#asahi-gpu · 2023-08-11 13:11
alyssalina: oh boy
#asahi-gpu · 2023-08-11 10:23
linaIt's what's causing one of the firefox issues
#asahi-gpu · 2023-08-11 10:23
linaI have this really weird shader bug which goes away if I add dummy code
#asahi-gpu · 2023-08-11 10:23
linaalyssa: I need your help ;;
#asahi-gpu · 2023-08-11 00:33
alyssathis is a.. weird bounds check
#asahi-gpu · 2023-08-11 00:33
alyssasaturating(offset + 4) > ((r0 << 2) - 32) ? ((r0 << 2) - 32) : offset
#asahi-gpu · 2023-08-11 00:14
alyssanini
#asahi-gpu · 2023-08-11 00:03
Maryokay will do, probably tomorrow as it's already 2AM oops
#asahi-gpu · 2023-08-11 00:01
alyssamove the call into the gallium driver (immediately before agx_preprocess_nir) and out of the compiler and then you should be fine
#asahi-gpu · 2023-08-11 00:01
alyssaMary: meh, the pass is probably technically broken but
#asahi-gpu · 2023-08-11 00:00
Maryso I wonder should I just add a nullcheck and upstream that? or just like disable that pass
#asahi-gpu · 2023-08-11 00:00
alyssathat's correct
#asahi-gpu · 2023-08-11 00:00
MaryAnd from what I could tell it seems to only be used in gallium in general...
#asahi-gpu · 2023-08-10 23:59
alyssaMary: spicy
#asahi-gpu · 2023-08-10 23:59
MarySo I have some crash on agxv related to "nir_lower_fragcolor" pass assuming a deref_var being present for the output... but that's not actually right for SSBO writes (for example with https://gist.github.com/marysaka/af8ef5e017bed5b2ca93ad8ff75bc701)
#asahi-gpu · 2023-08-10 00:28
alyssaexample how to teach kmsro about a new display controller: a1ed056ee6334ac58ce5e8120818fb88ba696f85
#asahi-gpu · 2023-08-10 00:27
alyssakmsro just tries all the gpus
#asahi-gpu · 2023-08-10 00:27
alyssamesa has a master list of known display controllers, as long as it's added to the list it should work
#asahi-gpu · 2023-08-10 00:26
alyssalina: wdym
#asahi-gpu · 2023-08-10 00:26
linaWe do have a dummy framebuffer but it's not probing, I think something is messed up with the power domains since these devices don't actually have disp0
#asahi-gpu · 2023-08-10 00:25
linaalyssa: The problem with kmsro is that it hardcodes what display controllers go with what drivers
#asahi-gpu · 2023-08-10 00:18
alyssaoh, I think vkms is the correct way to do this
#asahi-gpu · 2023-08-10 00:17
alyssa(did we ever get simpledrm + kmsro working?)
#asahi-gpu · 2023-08-10 00:17
alyssathe hacky way being m1n1 reserving a carveout of ram and pretending there's a simpledrm framebuffer there
#asahi-gpu · 2023-08-10 00:15
alyssa(Just enough to start weston or sway and run the "on screen" cts tests)
#asahi-gpu · 2023-08-10 00:14
alyssalina: surely there's some way to get a fake headless display DRM device?
#asahi-gpu · 2023-08-10 00:12
alyssaso either that's not the forwarding bit, or there's something really spooky happening
#asahi-gpu · 2023-08-10 00:11
alyssaNot sure how to reconcile these observations.
#asahi-gpu · 2023-08-10 00:11
alyssaie the bit that controls forwarding into the texture pipe
#asahi-gpu · 2023-08-10 00:10
alyssalabelled as 'forward'
#asahi-gpu · 2023-08-10 00:10
alyssaturns out we already handle that bit in applegpu
#asahi-gpu · 2023-08-10 00:10
alyssathat bit I identified as controlling a barrier with iterate instructions
#asahi-gpu · 2023-08-10 00:09
alyssadougall: ..and the plot thickens
#asahi-gpu · 2023-08-09 23:28
jannaulina: no luck with dcpext on t6020. While the dcp/iomfb protocol didn't change between 13.3 and 13.5, the dptxport protocol did change. not dramatically and I think I fixed everything. It seems to work but dcp fails detect a display
#asahi-gpu · 2023-08-09 20:27
alyssawe are underallocating. I do not know why. We don't seem to have much test coverage for block-compressed textures

plus ›